Prediction of sorption enhanced steam methane reforming products from machine learning based soft-sensor models

Nkulikiyinka, Paula; Yan, Yongliang; Güleç, Fatih; Manovic, Vasilije; Clough, Peter T.

Abstract

Carbon dioxide-abated hydrogen can be synthesised via various processes, one of which is sorption enhanced steam methane reforming (SE-SMR), which produces separated streams of high purity H2 and CO2. Properties of hydrogen and the sorbent material hinder the ability to rapidly upscale SE-SMR, therefore the use of artificial intelligence models is useful in order to assist scale up. Advantages of a data driven soft-sensor model over thermodynamic simulations, is the ability to obtain real time information dependent on actual process conditions. In this study, two soft sensor models have been developed and used to predict and estimate variables that would otherwise be difficult direct measured. Both artificial neural networks and the random forest models were developed as soft sensor prediction models. They were shown to provide good predictions for gas concentrations in the reformer and regenerator reactors of the SE-SMR process using temperature, pressure, steam to carbon ratio and sorbent to carbon ratio as input process features. Both models were very accurate with high R2 values, all above 98%. However, the random forest model was more precise in the predictions, with consistently higher R2 values and lower mean absolute error (0.002-0.014) compared to the neural network model (0.005-0.024).

Citation

Nkulikiyinka, P., Yan, Y., Güleç, F., Manovic, V., & Clough, P. T. (2020). Prediction of sorption enhanced steam methane reforming products from machine learning based soft-sensor models. Energy and AI, 2, Article 100037. https://doi.org/10.1016/j.egyai.2020.100037

• Journal Article Type: Article
• Acceptance Date: Nov 6, 2020
• Online Publication Date: Nov 11, 2020
• Publication Date: 2020-11
• Deposit Date: Jun 22, 2023
• Publicly Available Date: Jun 23, 2023
• Journal: Energy and AI
• Electronic ISSN: 2666-5468
• Publisher: Elsevier
• Peer Reviewed: Peer Reviewed
• Volume: 2
• Article Number: 100037
• DOI: https://doi.org/10.1016/j.egyai.2020.100037
• Keywords: Artificial Intelligence; General Energy; Engineering (miscellaneous)
• Public URL: https://nottingham-repository.worktribe.com/output/22182741
• Publisher URL: https://www.sciencedirect.com/science/article/pii/S2666546820300379?via%3Dihub